custom-agent-remediation-plan

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the jq utility within references/checks.md to perform structural and content validation on local JSON data. This is a controlled use of a common utility to ensure data integrity before findings are processed.
  • [PROMPT_INJECTION]: The skill ingests data from an external local file provided in the $PLAN environment variable, which represents a potential indirect injection surface. 1. Ingestion points: references/checks.md. 2. Boundary markers: The input is processed as structured data rather than free-form text. 3. Capability inventory: Use of jq for validation. 4. Sanitization: The skill implements multi-stage regex-based validation filters that reject placeholders and ensure data substantivity, mitigating risks of schema confusion or instruction injection.
  • [SAFE]: The skill incorporates comprehensive safety guardrails, including explicit requirements for audit logs, human escalation paths, and kill switches for any items classified as high-risk or action-capable. It also explicitly limits its own authority to avoid unauthorized production changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 12:13 PM
Security Audit — agent-trust-hub — custom-agent-remediation-plan